References to an unreleased OpenAI feature called "o," billed as an always-on assistant, briefly appeared in marketing material for the $100 ChatGPT Pro plan, BleepingComputer reported Sunday. OpenAI has not confirmed or denied the feature, and its capabilities are unknown, but configuration strings point to an assistant that may handle email.

What Happened

Users on X, including the account TestingCatalog, captured screenshots of a Pro plan benefits list that included "o" described as the subscriber’s always-on assistant. The listing then disappeared. The other items on the same list were more familiar: additional usage for Work and Codex, maximum memory, 100GB of file storage and early access to new features.

BleepingComputer also examined configuration references tied to the feature. They include a display_name value of "o" and an email_suffix value of "-o." The outlet read the second string as a hint that the assistant could be given its own email identity, or could be built to work with a user’s email, but OpenAI has said nothing about either possibility.

RelatedEight States Sue OpenAI as Iowa Seeks Court Halt on Testing of Advanced AI Models

The same reporting turned up references to an internal system called Aeon that is built around custom agents for ChatGPT workspaces. BleepingComputer said Aeon appears to be separate from the consumer-facing "o" feature.

Why It Matters

ChatGPT has mostly worked as a conversation that starts when the user types and stops when the reply arrives. An assistant marketed as always on implies something different: a process that keeps running between prompts, watching for work or acting on a schedule. If the email strings mean what they appear to, the assistant would also reach into one of the most sensitive data stores most people have, where it could read, sort or send messages.

That would raise practical questions OpenAI has not addressed, including what the assistant is allowed to do without asking, how users review its actions and whether an email address ending in "-o" would be visible to the people it writes to. Placing the feature in the $100 Pro tier would also make it a reason to pay for the higher plan rather than a free-tier feature.

Context

The leak lands in the middle of a contest over persistent personal agents. Meta used its Connect event last week to promote its Muse agent, including a pocket-sized Muse Charm device meant to keep the agent with users throughout the day. OpenAI, meanwhile, has been reshaping its paid tiers; BleepingComputer has also reported on a $500 ChatGPT Pro Max plan with faster Codex.

The leaked copy placed "o" alongside the plan’s existing perks rather than presenting it as a separate product. It did not include a launch date, a price change or any description of what the assistant can do.

What to Watch

OpenAI holds its DevDay 2026 event in San Francisco Tuesday, Sept. 29, the most likely venue for any announcement. The key questions are whether "o" ships under that name, whether it is limited to Pro subscribers, and whether it can read or send email on a user’s behalf. Any link between the consumer assistant and the Aeon agent system for workspaces would also show how OpenAI plans to split personal and business agents.